There are two variations: [Strong] and [Flexible]. The strong provides a strong representation of the style, but might struggle with complex prompts. The flexible trades some "likeness" with flexibility, suited for complex prompts (see first image).

Since Civitai lets us upload multiple files, I put both.

This lora is trained at 1024x and 1536x. Can achieve mid-res and high-res. Recommended resolutions for high-res are: 1216x1832, 1832x1216

Most of the dataset was gathered from Bakunyuu NTR party.